One message: the Inspector
Click an agent’s message to open it in the Inspector, on the right of the window. Turn shows whether the turn finished, how long it took, how many tokens went in and out, and what the room has spent today. Prompt makeup shows what the prompt was made of: tool schemas, system prompt, history and your message, each with its share of the total. See what was sent opens the agent’s Context tab with that turn’s prompt already open.When the total looks huge
Some agents run on a command-line tool (Grok, OpenCode) that works through several tool steps inside one turn. Each step sends the whole prompt again, and the provider reports the total for every step. A turn can show 2.2M input tokens when the prompt itself was about 31k. When that happens, the Inspector says Input, all steps for the total and shows One prompt beside it. To spend less on these agents, cut the number of tool steps or the size of the prompt.Everything the agent carries: the Context tab
Click an agent’s name and open Context.- Next turn shows how big the next prompt will be, against the model’s limit.
- This sitting shows turns, spend, and input split into new and cached.
- Sent prompts lists the latest prompts that went to the model. Open one to see what it was made of, then the system prompt, history, your message and tool schemas as they were sent. Very long parts are cut short in this view. Porto keeps the latest 48 and clears them when its services restart.
- System prompt lists its parts, largest first. Open any part to read it.
- Messages lists what the agent will see, newest first.